A Theoretical Attempt at Meta-Analysis

If you've ever played trading card games, then you'll have same familiarity with my current topic. I grew up playing all kinds of card games, but trading card games were always my favorite because of the diversity in how you could play them. You could be aggressive with quick, hard-hitting strategies, conservative with control and lock-down strategies, or clever with endless-looping combinations that "break the game" and result in an auto-win or a special end-game format to handle the "broken-game condition." The variety of different approaches results in a competitive state labelled as "Meta", where X% of the decks are Strategy A, Y% of the decks are Strategy B, Z% of the decks are Strategy C, and etc. For example, if you attended a tournament and expected X + Y > 60% of the meta, you would want to construct and play an anti-meta deck, which means your deck would beat "on average" more than 60% of the decks being played at the tournament. In more colloquial terms, it is akin to playing a "rock-type" deck when everyone else is playing a "scissors-type" deck (and hoping you don't randomly run into the less than 10% playing a "paper-type" deck).

College basketball demonstrates a meta-like quality, given the variety of offenses (passing motion, dribble-drive motion, swing, point-action reversal, shuffle, princeton) and defenses (pass-denial M2M, pack-line M2M, zone, and an occasional full-court havoc/hell press) in the game. What I am more interested in (and what this article will look at) is a statistical-based meta for the game and the tournament.

Meta of the Game

As much as I gripe about it on this blog, it goes without saying that the rules of the game dictate the meta of the game. It pains me even more that I can prove it (and have done so thoroughly on this blog).

This table shows the Four Factor national average in college basketball since 2008. The color-coding proves the impact of rule changes on the game. The blue coloring represents the years impacted by the institution of "Freedom of Movement" (and a friendly reminder that I still hate it with a passion). The yellow coloring represents the years impacted by the 30-second shot clock (formerly a 35-second shot clock for the non-yellow years). The beige coloring represents the years impacted by the 20'9" arc for 3-pointers (19'9" arc for the 2008 year in white). The year coloring of green, red or white indicates sanity, insanity, or neutral (respectively) according to various metrics published on this blog.

The impacts are noticeable.

• Freedom of movement (blue) has artificially boosted Adjusted Efficiency from a 100.5-101.7 range to a 102.0-104.2 range and has artificially exacerbated FT Rate from a narrow range of 36.0-37.8 to a volatile range of 33.1-40.6. As I have bluntly stated numerous times in the past, it has created a meta of artificially boosted points per possession by punishing defenses for playing defense.
• The 5-second reduction in the shot clock resulted in an artificially boosted Tempo of 69.1-69.5 possessions per game.from that of 64.9-67.3 possessions per game.
• The extension of the 3-point arc began with a lower 3-Pt% but as players accustomed themselves to it, the percentage returned to its previous levels (a phenomenon currently happening now with the extension of the 3-point arc to the current 22'1.75" arc).

It suffices to say that the current meta in college basketball is brain-dead dribble-drive basketball, launch a lot of 3-point shots, and Toro-style defense where the defender lets the ball-handler through to the rim to avoid a foul call like a bull-fighter lets the bull drive through his/her red cape. Also obvious is the lessening importance of the Factors not targeted by the rule changes like turnover percentage and offensive rebounding percentage. If dribbling-based offenses are the meta, it's less likely that turnovers will occur since the majority of turnovers are typically from passing the ball. If dribbling collapses the team-defense toward the rim and offenses keep non-dribblers stationed on the perimeter, offensive rebounding should also decline since more defenders are near the potential defensive rebound. This brand of basketball is very primitive in its nature and very pedestrian to watch (which is why I don't watch the NBA nor do I want college basketball to be the NBA pre-school).

This chart portrays the continuing trend of the college basketball meta through 2021.

Meta of the Tournament

Since four of the years (2016-2019) on the table have the same meta, I thought I would further investigate the "meta concept" as it applies to the tournament and the teams competing in it (much like the variety of decks competing in a trading card tournament). Below is two tables of the Four Factor averages of Quality Curve teams and 1-12 seeded teams.

I'm simply looking for anomalies in one or two Factors and seeing if these anomalies produce a rock-type approach in a scissors-type tournament.

• 2019: 35.3%/35.4% for 3P% and 32.6%/32.5% 3P%D (look for high 3P% to combat this below-average meta), 113.4/113.3 AdjOE and 93.7 Adj Eff (look for low 2P%D to combat easy points)
  
• 2018: 48.5%/48.4% EFGD% (look for high 2P% to exploit this below-average meta) and 69.2/69.4 Tempo (look for low TOR because higher possession counts with fewer turnovers equals higher shot volume)
• 2017: 37.5%/37.4% for 3P% (look for low 3P%D to combat this above-average meta or look for high OR% to equalize it with high shot volume and high percentage looks at the rim). NOTE: The relative comparisons for 2017 are unmatched by any of the other three years, and I don't know how I would begin to evaluate this meta.
• 2016: 52.4%/52.3% for EFG%, 47.2%/47.3% for EFGD%, and 51.2/50.9% for 2P% (look for high 2P% and high 3P% to combat this below-average meta or look for low TOR and high ORB to combat poor shooting with more shots per possession.

Let's take a look at how each of these meta-predictions play out by comparing them to teams that advanced to at least the Elite 8. I won't detail each one, but you can judge the results for yourself.

2019: Look for teams with high 3P% and low 2P%D

2018: Look for teams with high 2P% and low TOR

2017: Look for teams with low 3P%D or high OR%

2016: Look for high 2P% and high 3P% or low TOR and high ORB

From the results above, I do like the potential of meta-analysis. Instead of looking at how one team matches up against another, meta-analysis looks at how a team matches up against the entire field in order to predict a team's chances to make a deep run. There are several outliers in each year, and these outliers may have more to do with the meta of their pod or region than the meta of the whole field (and this would be very good extension of this type of analysis). The four years in question are comparable because they all fit under one unique meta (20'9" 3-point arc, Freedom of Movement, and 30-second shot clock). No other year has these three exact modalities. I'm not sure how comparable 2021 is to the four years above due to the change in the 3-point arc distance. The only year that is meta-comparable to 2021 is 2020, but there are no end-results to make comparisons since the 2020 tournament was cancelled. The frequency of change in the meta will determine its usability as a predictive model. If gameplay rules keep changing every three to four years, then the meta analysis model will only be useful in years 3 and 4 of that specific meta so that anomalies can be identified and the key factors sort out the rocks from all of the scissors. Even if the model isn't usable in its first few years of a new meta, the averages for the QC and 1-12 seeds can be used as thresholds for champion and final four contenders/pretenders (a model made popular by my mentor). It's another analytical tool that my wandering mind has envisioned, and I hope to put it to good use in a few years (maybe even this year as a high-risk model). Until then, thanks for reading my work and I hope to see you around Bracket Crunch Week.